SIGCHI Conference Paper Format

1 downloads 0 Views 710KB Size Report
ABSTRACT. The recent success of Nintendo's Wii and multi-touch input .... Nintendo is market-leader ..... Switch to next image using a waving gesture from left to.
Designing a multi-purpose capacitive proximity sensing input device Andreas Braun Fraunhofer IGD Fraunhoferstr. 5, Darmstadt, Germany [email protected] +49 6151 155 208

Pascal Hamisu Fraunhofer IGD Fraunhoferstr. 5, Darmstadt, Germany [email protected] +49 6151 155 579 computer interfaces based on capacitive proximity sensors. However they are mostly limited to a few supported hand gestures [16] or only support applications [15].

ABSTRACT The recent success of Nintendo’s Wii and multi-touch input devices like the Apple iPhone clearly shows that people are more willing to accept new input device-technologies based on intuitive forms of interaction. Gesture-based input is thus becoming important and even relevant in specific application scenarios. A sensor type especially suited for natural gesture recognition is the capacitive proximity sensor that allows the detection of objects without any physical contact. In this paper we extend the input device taxonomy by Card et al to include this detector category and allow modeling of devices based on advanced sensor units that involve data processing. We have created a prototype based on this modeling and evaluated its use regarding several application scenarios, where such a device might be useful. The focus of this evaluation was to determine the suitability of the device for different interaction paradigms.

In previous work we have followed a similar approach, using a single sensor [2] in order to control simple demonstration applications. Our focus for this paper was to go a step further and create a generic input device based on an array of capacitive proximity sensors that should be able to track the location of the user’s hand. The idea was to create a multi-purpose device that can be used in various application scenarios. Since we wanted to go beyond previous technology-driven attempts we were looking for a way to formally model such devices and integrate them into existing input device taxonomies. Therefore we have evaluated several classical approaches focusing on their ability to allow the modeling of novel input devices. Based on this evaluation we decided to extend the modeling system and taxonomy of Card et al [5]. This includes the specific addition of capacitive proximity sensors but also a general methodology to simply modeling of modern input devices that rely on data pre-processing, which is difficult or even impossible to describe with the existing method.

Categories and Subject Descriptors B.4.2 [Input/Output and Data Communications]: Input/Output Devices - Data terminals and printers. H.5.2 [Information Interfaces and Presentation]: User Interfaces - Evaluation/methodology, Graphical user interfaces, Input devices and strategies, Interaction styles.

General Terms Design, Verification.

Keywords Input devices, interaction design, modeling, evaluation.

1.

INTRODUCTION

Natural interaction has been a strong research focus in human computer interaction the last few years. The capacitive proximity sensor is a detector able to sense the presence of objects without physical contact over a distance ranging from a few centimeters up to several meters, depending on sensing system and setup. This characteristic is allowing this detector type to recognize spatial gestures performed by a user.

Figure 1. Conceptual rendering of hand interacting with the active area of hand-tracking input device Using these additions we have formally modeled a hand-tracking prototype based on an array of capacitive proximity sensors. Various pre-processing algorithms generating high-level information from raw sensor data have been tested on this prototype, in order to evaluate the functionality of such a device in several usage scenarios.

There have been various attempts on creating non-haptic human-

Unlike mechanical input devices the user receives no haptic feedback from this device; there is no feedback when entering the interaction area above the device. In addition there is also no direct visual feedback from the interaction area, where the user is directly touching the items he wants to manipulate, as in touch-

Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. PETRA 2011, May 25–27, 2011, Heraklion, Crete, Greece. Copyright 2011 ACM 978-1-60558-246-7/08/04…$5.00

1



screens or tangible interfaces. Despite those challenges we have determined various application scenarios that can profit from such an input device:   



Unobtrusive media center control Medical application in sterile environments Interaction device for the physically challenged

 

In the final part of the paper we evaluate each of those scenarios using different demonstration applications with distinct interaction paradigms and interfaces:   

Table 1. Evaluation of input device taxonomies Criteria

Gesture-controlled image viewer Menu-based, movement-controlled 3D object viewer Position-based regional image & document zoom

Generality

RELATED WORK

2.1. Natural Interaction

Extensibility

Diverse research projects of the last decade have investigated the next evolutionary steps in HCI, trying to achieve a more natural interaction between user and machine, i.e. through gestures, speech and physical interaction [13]. Popular approaches include microphone arrays to recognize voice and locate the user [12], multiple video cameras using feature extraction to identify user movements [15] and large active display areas capable of identifying user touches [10]. The of gesture control schemes in gaming is a very notable achievement. Nintendo is market-leader with the Wii console. Microsoft (Kinect) and Sony (Move) are bringing similar systems to market in fall 2010. Another very successful gestural input device is the Apple iPhone that has been influencing user interface and interaction design in mobile multimedia devices ever since its introduction in 2007.

No audio, vision If in six defined categories

Buxton Only continuous handoperated devices No foot, audio, vision If continuous handoperated device Supported devices well-defined

Card Composition Virtual input devices

No audio, vision If input, output domains can be defined

Multiple Device-based categorization modeling Definiteness for single devices The taxonomy of Foley et al was discarded early, as the strict categorization makes it difficult to integrate devices based on novel interaction forms, like gestural input. Another interfering factor is the ambiguous categorization of multi-purpose devices. Buxton et al allows for good categorization but is not well-suited for versatile sensors like capacitive proximity sensors that can track a hand, but also feet or the whole body. Also multi-purpose devices that may be configured for various degrees-of-freedom would lead to ambiguous categorization. In conclusion the taxonomy and modeling by Card et al was chosen as candidate to integrate the new sensors and input devices. It allows for modeling various systems based on the same sensor platform by using a different composition. However the main advantage for our purposes is the virtual input device that can be modified to provide an elegant way to model complex data pre-processing, which will be shown later in this section.

2.2. Input Taxonomies The growing number of input devices has led to attempts in categorization of those, creating various taxonomies of input. An early user-centric system by Foley et al [7] is based on the user’s intentions that have been grouped into six categories of generic transactions. Buxton et al [3] used an approach restricted to hand operated devices that tries to categorize devices based on features of the human sensory system and provided degrees of freedom. A final classical approach is modeling the physical properties of each sensor attached to the input device and combines them using several composition operators. This system has been developed by Card et al [5].

In order to keep this document self-contained, the modeling system of Card et al is described briefly using an example. The approach is to model all input devices from singular input devices, usually sensors determining a certain physical property. These singular devices are described using a six-tuple 〈 〉, consisting of a manipulation operator M, the input domain In, system state s, a resolution function R, transforming the input domain into output domain Out and a final attribute W that acts as placeholder for additional functionality. The x-Movement wheel in a mechanical mouse has the six-tuple 〈 〉. A change in x-direction is measured as a real number, transferred using a linear function, leading to a real number output and no additional information.

2.3. Capacitive Proximity Sensors Capacitive proximity sensing is a rather old application that first appeared in the 1920s [8]. Throughout the 1990s it has been a research interest at the MIT Media Lab [11]. Research of those systems is ongoing, with several commercial and scientific systems available [6], [4].

3.

Foley No combined devices

Flexibility

We have done the initial evaluation by gathering qualitative feedback from expert users that are experienced in the usage of various input devices

2.

Flexibility: Adjustability towards layout changes in devices (e.g. traditional mouse vs. scroll wheel mice) Generality: The scope of included input devices. The majority of input devices should be included. Extensibility: Simplicity of adapting the taxonomy regarding new device categories. Definiteness: Ambiguous device categorization should be avoided.

INPUT DEVICE MODELING

While being an established sensing device, it has not yet been tried to integrate input devices based on capacitive proximity sensors into common input device modeling systems and taxonomies. We have evaluated the three systems mentioned earlier regarding the following criteria:

2

Figure 4. Modeling modes of capacitive proximity sensors. Left proximity mode - right state mode We consequently use a simplified model, whereas the sensor is modeled as a point-shaped, isotropic proximity detector that does not detect an absolute distance value, but proximity relative to its maximum detection distance. In order to allow for more diverse input devices we model two different sensor modes. The first is called proximity mode and is producing values directly proportional to the detected capacitance. Dependent on firmware and resolution this might be a 2-byte integer value, where higher value signifies the detected object is closer to the sensor, and 0 refers to no object within detection range. The second mode is called state mode. This is basically a sensor-internal quantization of the values into n different states, which can be used to model capacitive buttons that are commercially available and applied in various devices. Both modes are visualized in Figure 4.

Figure 2. Compositional features of the Card et al taxonomy Figure 2 displays the compositional features of the taxonomy by showing the example of a mechanical mouse. The inputs of xdirection and y-direction sensors are merged into a xy-detector. Several other sensors, like buttons and scroll-wheel are added to the same device, leading to the multi-function devices that are very common today. Finally the device is connected to a virtual input device, the mouse cursor, prevalent in many operating systems to control graphical user interfaces.

We propose the following singular input devices and their associated six-tuples in Table 2 to model capacitive proximity sensors in both modes described. Table 2. Six-tuples according to Card taxonomy for both proximity sensor modes M

R Out W Transfer Real Proximity S Function Number Mode Static 〈 〉 S 〈 〉 State Mode Function The manipulation operator M is a position in spherical coordinates, relative to the position of the electrode. A single sensor is not able to detect the direction of the approaching object, thus determining angles requires input from multiple singular sensors.

Figure 3. Principle of capacitive proximity sensing

3.1. Capacitive proximity sensors The principle of a capacitive proximity sensor is to measure the change in capacitance of an electrode, caused by objects passing through the generated field. A simplified version of the principle is displayed in Figure 3. An electrode is periodically charged and discharged. If a conductive object passes through the electric field the capacitance of the system increases. By measuring the charge flowing back from the electrode the relative change in capacitance can be detected. Assuming the conductive object is of a uniform size, this change is roughly proportional to the distance between electrode and object, following the law of capacitance between parallel plates.

In Real Number

s

3.2. Modeling complex data processing Many modern input devices include microprocessors that allow for more or less complex data pre-processing before any events are sent to the actual computer. One example of such a system is the optical mouse that uses optical flow algorithms to detect movement and is send calculated xy-movement to the attached computer [9].

The actual electrical properties of the sensing system are depending on electrode geometry, the detected object geometry, as well as various environmental factors like air humidity or disturbance from nearby electrical circuits. Solving such a system using electrostatics is highly complex and not practicable [1].

Figure 5. Data processing virtual input device connected to physical input device output and computer Our idea was to integrate complex data processing into the Card taxonomy by encapsulating it into a virtual input device, placed between physical input device and computer (Figure 5). It

3

Resulting planar hand location ( ̅ , ̅) is calculated using the sums over sensor positions (xi, yi) and sensor values vi as weight. While there are different, more advanced localization algorithms, this approach has several advantages: The implementation is easy and computationally inexpensive, thus applicable on a low-power microcontroller. Furthermore it requires only a single sensor to work, different to other common approaches, including trilateration or multilateration that require at least three active receivers. For many hand positions, especially at the edge of the device less than three sensors detect an object.

transforms data generated by all singular or merged input devices into new output, exceeding the scope of the resolution function described in the original version. Incoming data is put into a data structure we call Fusion Structure and transformed using various algorithms resulting in an output data structure. We suggest the notation in Figure 6 for a more detailed description of the data processing virtual input device.

Figure 6. Data processing virtual input device detailed The input device is illustrated as grey box that shows applied algorithms and some additional information, without describing the process in full detail.

3.3. Proximity box modeling Designing our input device we tried to meet the following requirements:   

Suitability for various application scenarios Easy accessibility for setup changes Balance between cost technical sophistication

Figure 8. Piecewise linear interpolation used to determine zposition of detected object Since all sensor units are located on the same plane, a different approach is required to determine the z-position. The idea is to simply use the sum of all sensor values and associate it to height. Since sensor values increase when an object is approaching, this assumption is viable if the sensor distribution on the plane is sufficiently dense. Since the sum of values is increasing in nonlinear fashion, we decided to use piecewise linear interpolation between fixed threshold levels, as seen in Figure 8, which delivered satisfying results.

Therefore we decided on a portable device housing six sensor units. Tests have shown that this number of detectors is providing a good balance between cost and functionality. Thus we model the device from six capacitive proximity sensing singular input devices and a single data processing virtual input device. Figure 7 shows the composition model according to the extended Card model. The six capacitive proximity sensors are merged to a single sensor array that is sharing a layout. Each sensor is assigned a specific position in the array. The combined output is connected to the data processing virtual input device.

Figure 9. Data processing virtual input device for modeled input device Finally we noticed that the sensor devices are suffering from a considerable amount of noise, which required us to apply some interpolation on detected locations, to allow for smooth interaction. Two different instances of linear interpolation are running for planar position and height. The process results in the virtual data processing input device depicted in Figure 9, based on the notation detailed earlier.

Figure 7. Proximity box composition according to extend Card model As described earlier the data processing virtual input device is transforming the input located in the fusion structure into an output structure that is sent to the next device. The fusion structure is composed of six sensor data structures, containing the sensor value, a unique sensor ID and sensor location in xy-coordinates.

4.

The additional positioning information is required to determine the hand position. The calculation uses different approaches for xyposition and z-position. In order to determine object location relative to the plain we use a weighted average algorithm: ̅

∑ ∑

,

̅

PROTOTYPE

In order to test the various scenarios we needed a prototype that enables easy setup changes, while remaining portable. Therefore we decided to use sensors that are transmitting wireless to a base hub, and use wire as electrodes to experiment with different configurations. The disadvantage is that a dedicated power source is required, in case of our first prototype AAA batteries.

∑ ∑

After scanning the market for available solutions we decided to base our prototype CY3271 PSoC First Touch Starter Kit by

4

Cypress Semiconductor that combines capacitive proximity sensing capabilities with wireless data transmission using a proprietary 2,4GHz protocol. Coiling the wire and placing it on acrylic tubes to get closer to the surface provided to be a good configuration with a good detection range.

Figure 11. Gesture-controlled image viewer demonstration application A typical application for media centers is the display of images in a slideshow. Therefore we decided to implement this into our demonstration software with some common features enabled. As can be seen in Figure 11 the current position of the detected hand is depicted as green circle, with the path of previous locations displayed in fading circles providing instant visual feedback, thus allowing the user to perform supported gestures more easily.

Figure 10. Conceptual renderings and internal views of first (used for evaluation) and second prototype

The software iteration used is supporting four gestures:

This was the prototype used in our evaluation. However recently we have switched to a new iteration based on the same sensor type, yet with a more pleasing appearance and various modifications, like using a rechargeable battery for all sensors combined and switching to copper foil electrodes that provide a slightly improved signal-to-noise ratio. Both prototypes are displayed in Figure 10.

5.

   

DEMONSTRATION APPLICATIONS

Regarding the general design of demonstration applications, the visual feedback is very important for this type of device. The lack of haptic feedback and direct visual presentation otherwise can make interaction difficult. All applications have therefore been developed using the XNA framework as overlay on Direct3D to freely design the user interfaces without operating system restrictions.

Switch to next image using a waving gesture from left to right Switch to previous image using a waving gesture from right to left Zooming into the picture if movement from bottom to top is detected Zooming out of the picture on movement from top to bottom

The gestures have two different modalities. The waving gestures are looking for a certain threshold to be exceeded and fire an event on occurrence. The moving gestures fire events until direction is changed or movement has stopped. Recently we also have added a video viewer to the software suite that is using the same gestures for pausing and restarting the video and controlling volume instead of zooming.

As described earlier we have determined three application scenarios in which our modeled input device shall be tested.

5.2. Medical applications environment

5.1. Unobtrusive media center control

in

sterile

The rise of medical image acquisition technologies have led to huge advances in modern hospitals, allowing more complex surgeries and increased chances for recovery. Modern MRT or PET devices deliver highly detailed 3D representations that are useful in planning the surgery but also occasionally have to be consulted in the ongoing procedure.

Modern living rooms often feature many single-purpose devices, usually steered by individual infrared remote controls that have different layouts and often lead to confusion, especially for elderly users. In recent years media centers have appeared as alternative. These are devices based on PC hardware and powerful operating systems that try to provide access to all sources of media through a single interface. However interaction is still based on complex remote controls or keyboard and mouse.

The benefit of using capacitive proximity sensing in this context is sterility. The surgeon is not allowed to touch any surface while operating on the patient to prevent infection. Thus if he is required to consult the 3D images taken before the operation he has to rely on an assistant which might lead to delays. A gestural interface based on proximity sensing would allow the surgeon to directly manipulate the images without having to risk sterility.

Our vision for capacitive proximity measuring input devices in this context is the unobtrusive integration into living room furniture. The sensors work through any non-conductive material, thus are suited to be integrated into wooden or stone/ceramic furniture if the upper layer is not too thick. Our prototypes work with a rather thin acrylic layer on top; however tests have shown that it also works through wooden tables with a thickness of approximately 4cm.

5

Figure 13. Screen lens image & document viewer in different zoom states Our demonstration application in this scenario implements a screen lens for text and images. A regular feature in many operating systems that is suited for the elderly or otherwise visually impaired persons. We use it to demonstrate a direct location-based interaction. A regional zoom is applied at the current detected xy-position. The z-position determines the zoom factor; touching the prototype will lead to the highest magnification. Switching between documents or images is implemented using active regions. If the hand is held in the right region for a certain time the software switches to the next time. Likewise remaining on the left side will load the previous items.

Figure 12. 3D object manipulation showing an image of a vertebral column Therefore our demonstration for this application scenario is a 3D object viewer featuring a menu-based interface that is controlled by a combination of emulated point-and-click, as well as movement control. The common pan, rotate and zoom features of 3D object viewers are supported:   

Pan is a direct transposition of the hand movement onto the scene Rotate is creating an axis vertical to movement through the object central point and rotates relative to movement speed Zoom is working analogue to the zooming in the image viewer demonstration application

6.

EVALUATION

The evaluation was done with 18 persons, 12 male and 6 female with an age range between 20 and 50. Participants were invited from our research institute and nearby university and can be described as proficient in PC usage. We selected a group acquired with modern HCI technology, partially even working in UI research. The target of this evaluation was to get expert feedback, in order to optimize hard- and software as well as determining future steps in expanding this technology. Thus we associate the expected feedback to two groups:

In order to increase the available screen estate the menu is initially hidden and has to be activated by moving the hand to the left side of the input device. Menu items are activated by holding the position on the item for a given amount of time. One additional feature allows switching between several preloaded models via menu entry.

 

5.3. Interaction device for the physically challenged

General performance of the device Comparing the different application scenarios and interaction paradigms

Each participant was given a five minute introduction into the features and interaction paradigms of each demonstration application. Afterwards they had ten minutes to test the program and later fill out a questionnaire about their experience. Besides acquiring general information the participants had to grade precision, how intuitive it is to use and if they could expect using it for long term purposes. Grades are ranging between 1 (worst) and 10 (best).

Certain diseases and health related problems may lead to difficulties in using common input devices like mouse and keyboard. Examples include carpal tunnel but also amputated hands or forearms. There are several input specialized input devices that target single aspects of this group. Capacitive proximity sensor arrays can be applied very flexible. Changes in electrode design can lead to device with a geometry differing from our prototype. If that is combined with adjustments in gesture recognition algorithms it is possible to create devices based on the same technology that are suited for many different symptoms and diseases.

The results are displayed color-coded in the following bar chart, e.g. two participants graded precision with a 2, while three graded it with an 8.

6

device itself. Many of the negative aspects can be attributed to technical deficiencies of our prototype. Precision and interaction speed are the main concerns the participants had. We are optimistic that we can improve these factors by using different sensor and data transmission technologies and optimized antenna designs. The most important negative aspect noticed in the evaluation is the tiring of the arm after using the device for prolonged periods. This factor is also affecting prospects for long-term usage negatively. We realize that this is a negative aspect inherent to the system that has to be taken into account when designing applications and user interfaces. The input device is suited best for scenarios where short interaction is followed by longer breaks, e.g. media center applications or presentation software. If that is not the case the user interface has to be designed in a way that the user does not need to interact for long periods. A first addition to our software made after the evaluation was to modify the screen lens application, where this prolonged interaction is most prevalent. We added what we call sticky lens; that is that the software is detecting when a user stays in a position and height for a certain time and sets the lens as stuck, meaning the lens parameters are stored and the user can remove his hand. As soon as the hand enters again in different position the lens is movable again.

Figure 14. Evaluation results of device in general The precision of the device did get the lowest grading. Participants could imagine using such a device for longer times, depending on the application scenario. Participants mostly liked using our prototype to control applications and managed to learn the required gestures in a short amount of time. Regarding each application they had to grade their satisfaction analogue to general questioning between 1 (worst) and 10 (best).

A prevalent positive aspect in the evaluation is the intuitive use of the device. Provided with proper visual feedback, users were able to control the applications on their own quickly. They also considered such an interaction enjoyable due to its novelty. Similar reasons lead to the success of other gestural interfaces in the past few years. Based on this it is viable to further explore possibilities and limitation of input devices based on capacitive proximity sensing.

7.

CONCLUSION

In this paper we have explored the design of input devices based on capacitive proximity sensors. We have presented a method to formally model input devices based on capacitive proximity sensors. In order to achieve this we have added a new singular input device to the modeling system of Card et al and developed a method to integrate the complex pre-processing that is occurring in many modern input devices. Figure 15. Evaluation results of individual applications

To demonstrate this modeling we have formally described an input device based on six sensors that is able to detect the position of the user’s hand moving above it in three dimensions. In order to explore the performance and usability of this device we have built a prototype and created several demonstration applications showcasing typical features in medical context, media center control and interaction for persons with disabilities.

The results have shown that the general satisfaction of the participants regarding the demonstration applications has been fairly high, with a clear difference showing between the 3D object manipulation and gesture-based image viewer, screen lens on the other side. The positive remark we gathered can be roughly put into three groups:   

Finally we have done an initial evaluation with a group of expert users to explore the viability of our application scenarios and prototype, as well as getting feedback and ideas for future improvements.

The device very intuitive to use The idea of interacting this way is novel and interesting It is easy to control applications with those gestures

The evaluation results suffered from technical deficiencies in the prototype but also offered a positive outlook on the potential acceptance and future of this category of non-haptic input devices.

Likewise we identified three main groups for negative comments about the prototype:   

The device is not very precise The interaction speed is slow It can be tiring for the arm

Based on the evaluation results we have started to optimize hardware and software, resulting in a new prototype and additional features in our application suite.

6.1. Interpretation of results

8.

Evaluation results on the general questions as well as analyzing the comments lead to an overall positive picture of the input

We are trying to further explore various aspects of the technology and user interface design for this type of device in future research. 7

FUTURE WORK

We are currently working on a new iteration of the prototype that should have a much higher precision and increased interaction speed. Furthermore we want do explore user interfaces that are specifically designed for non-haptic interaction, supporting more gestures and having the ability to switch between various applications without having to fall back to traditional input devices like mouse and keyboard. In this context we also want to further explore the specific challenges of this interaction paradigm. E.g. we have noticed that it is sometimes difficult for a user to determine if his hand is currently in detection distance. Various approaches exist to determine a user’s intent that we want to explore. Finally we have also planned another evaluation, in which we want to compare the performance of several 3D input devices.

9.

[7] Foley J.D, Wallace V.L., Chan P., The human factors of computer graphics interaction techniques, In IEEE Computer Graphics and Applications, vol. 4, no. 11, IEEE Computer Society Press (1980), 13-48. [8] Glinsky A., Theremin: ether music and espionage, University of Illinois Press (2000). [9] Lyon R.F., The optical mouse, and an architectural methodology for smart digital sensors, In VLSI Systems and Computations, 1981. [10] Rekimoto J., Matsushita N., HoloWall: designing a finger, hand, body, and object sensitive wall. In Proc. of the 10th annual ACM symposium on User interface software and technology, ACM (1997), 209-210.

ACKNOWLEDGMENTS

[11] Smith J.R., Electric field imaging, Massachusetts Institute of Technology (1999).

We would like to thank all volunteers from the TU Darmstadt and Fraunhofer IGD Darmstadt that have been participating in our evaluation and provided helpful comments that will help us planning our future steps in researching novel interaction technologies.

[12] Sturim D., Silverman H., Brandstein M., Tracking Multiple Talkers Using Microphone-Array Measurements. In Proc. of the 1997 IEEE International Conference on Acoustics, Speech, and Signal Processing, IEEE Computer Society (1997), 371.

10. REFERENCES

[13] Valli A., The design of natural interaction, Multimedia Tools and Applications 38, 3 (2008), 295-305.

[1] Baxter L.K., Capacitive Sensors, John Wiley and Sons, 1996. [2] Braun A., Hamisu P., Using the human body field as a medium for natural interaction. In Proc. of the 2nd International Conference on PErvasive Technologies Related to Assistive Environments 2009, ACM (2009), No. 50.

[14] Wimmer R., Schmidt A., Holleis P., Kranz M., Thracker – Using Capacitive Sensing for Gesture Recognition. In Proc. of the 26th IEEE International ConferenceWorkshops on Distributed Computing Systems 2006, IEEE Computer Society (2006), 64.

[3] Buxton W., Lexical and pragmatic considerations of input structures. In ACM SIGGRAPH Computer Graphics, vol. 17, no. 1, ACM (1983), 31-37.

[15] Wren C.R. et al, Perceptive Spaces for Performance and Entertainment: Untethered Interaction using Computer Vision and Audition. In Applied Artificial Intelligence, vol. 11, no. 4, Taylor & Francis Group 1997, 267-284.

[4] CapToolKit, http://www.capsense.org. [5] Card S.K., MacKinlay J.D., Robertson G.G., The design space of input devices. In Proc. of the SIGCHI conference on Human factors in computing systems: Empowering people 1990, ACM (1990), 117-124.

[16] Zimmerman T.G., Paradiso J.A., Smith J.R., Allport D. and Gershenfeld N., Applying Electric Field Sensing to Human Computer Interfaces. In Proc. SIGCHI conference on Human factors in computing systems 1995, ACM Press (1995), 280287.

[6] Cypress CapSense, http://www.cypress.com/?id=1575&source=header.

8